One-stop logistics information intelligent retrieval and prediction method based on artificial intelligence

Abstract

The invention relates to a one-stop logistics information intelligent retrieval and prediction method based on artificial intelligence, and relates to the technical field of intelligent logistics. The method comprises the following steps: receiving a logistics demand input by a user in a natural language or a structure through AI dialogue interaction, and analyzing key information by using an intention recognition and entity extraction technology; the optimal transportation route is intelligently matched, the delivery warehouse is recommended, and the freight cost is dynamically calculated; comprehensively using a time sequence prediction model to analyze historical data so as to estimate the time efficiency of each link, evaluating the line service level, and generating a visual statistical report; and automatically detecting traffic restriction policies and providing detouring suggestions, and finally predicting key time nodes of the whole transportation process through AI simulation. The method effectively solves the problems that logistics information is dispersed, depends on manpower and is inaccurate in prediction, achieves efficient and accurate one-stop logistics decision support, and remarkably improves the efficiency and the intelligent level of logistics operation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent logistics and artificial intelligence, in particular to a one-stop logistics information intelligent retrieval and prediction method based on artificial intelligence, which is used for realizing integrated query, intelligent recommendation and multi-dimensional prediction of logistics information. BACKGROUND

[0002] At present, in the process of obtaining logistics information before goods are shipped, relevant personnel of an enterprise usually need to manually collect and integrate data across multiple independent links or systems, such as querying transportation routes, warehouse status, cost standards and policy requirements, etc. This method is tedious and time-consuming, and due to the problems of scattered information, different formats, and time difference, it is easy to miss or distort information, which seriously affects the accuracy and timeliness of decision-making.

[0003] On the other hand, the enterprise lacks a systematic monitoring and evaluation mechanism for internal line transportation service level, and it is difficult to obtain multi-dimensional data such as historical timeliness performance, service quality indicators (KPI), customer feedback, etc. This lack of data makes it difficult for the enterprise to establish a scientific prediction model and to make reliable transportation path planning and resource scheduling.

[0004] In addition, existing logistics cost estimation relies mainly on manual experience or static price tables, without fully considering external variables such as line dynamic changes, seasonal factors, policy adjustments, etc., resulting in significant deviations between estimated results and actual costs, with limited reference value. This not only makes budget control difficult, but also causes a lot of unnecessary time consumption in repeated confirmation and correction.

[0005] The reason is that there is a lack of a unified platform in the existing technology that can integrate multi-source data and has intelligent analysis and prediction capabilities. The information island phenomenon is serious, data processing relies on manual work, and algorithm support is insufficient, which cannot realize efficient and accurate logistics simulation and decision support.

[0006] The above method has the problems of scattered information, different formats, and poor timeliness, resulting in low decision-making efficiency and high error rate. In addition, the lack of systematic evaluation and prediction of transportation line service level makes it difficult for enterprises to plan paths, schedule resources, and control costs. There is no logistics information platform in the existing technology that can integrate multi-source data, support natural language interaction, and have full-process intelligent prediction capabilities. Therefore, there is an urgent need for an integrated solution that can deeply integrate internal and external data and realize intelligent retrieval and prediction through artificial intelligence technology. SUMMARY

[0007] In view of the deficiencies of the prior art, the present application aims to provide an intelligent retrieval and prediction method for one-stop logistics information based on artificial intelligence, which can deeply fuse and analyze the internal and external logistics data of an enterprise, output high-reliability prediction results through an intelligent model, thereby effectively assisting the enterprise in forward-looking planning and decision-making, and comprehensively improving the efficiency and economy of logistics operation, so as to solve the problems of information silos, high artificial dependence and low prediction accuracy in the prior art.

[0008] The above application object of the present application is realized by the following technical scheme:

[0009] An intelligent retrieval and prediction method for one-stop logistics information based on artificial intelligence, comprising the following steps:

[0010] Step S1, identifying key information in user input, and automatically matching line information meeting conditions maintained in the system;

[0011] Step S2, intelligently screening and recommending a suitable delivery warehouse through dialogue interaction;

[0012] Step S3, calling a preset price library, automatically calculating transportation unit price, special fees and total transportation cost, and feeding back in a natural language manner;

[0013] Step S4, intelligently analyzing historical loading, transportation and unloading time efficiency data based on the matched line information, and generating an overall transportation time consumption estimate;

[0014] Step S5, analyzing historical KPI indicators and service evaluation data of the line through AI;

[0015] Step S6, automatically querying and generating a visual report of statistical information such as historical cargo shipment tonnage and shipment number;

[0016] Step S7, automatically detecting the limit policies of the sending and receiving areas and providing detour suggestions;

[0017] Step S8, simulating and predicting each key time node in the transportation process, such as order receiving, dispatching completion, pickup completion, in-transit transportation, arrival at the destination, and unloading and signing, and providing full-process prediction instructions to the user in a dialogue manner.

[0018] As a further technical scheme of the present application, the step S1 of identifying key information in user input comprises the following steps:

[0019] User intent recognition is performed using a natural language understanding model to determine that the user intent is logistics query and prediction;

[0020] Key entities including destination, shipment tonnage, time information, cargo type and service requirements are extracted from user input using a named entity recognition model.

[0021] Map and standardize the key entities with the master database inside the system to generate structured query conditions.

[0022] As a further technical solution of the present application: the step S2 of intelligently filtering and recommending suitable delivery warehouses includes the following steps:

[0023] According to the user-specified departure location information, filter out available warehouses from the warehouse master data;

[0024] Based on a neural network recommendation model, calculate a recommendation score for each available warehouse, which comprehensively considers multiple dynamic and static factors such as real-time warehouse capacity, historical delivery time efficiency, current transportation capacity resource tightness, and transportation cost;

[0025] Output one or more warehouses with the highest recommendation score and their recommendation reasons.

[0026] As a further technical solution of the present application: the step S3 of calculating transportation costs includes the following steps:

[0027] According to the matched route and cargo information, query the base transportation price per unit in the preset price library;

[0028] According to service requirements, calculate one or more special fees through a rule engine, including express fee, unloading fee, road and bridge fee, or fuel surcharge;

[0029] Calculate the total transportation cost according to the formula total cost = (base unit price * billing weight) + sum of all additional fees.

[0030] As a further technical solution of the present application: the step S4 of generating an overall transportation time consumption estimate includes the following steps:

[0031] Extract historical time efficiency data for the matched route from the historical shipping order database;

[0032] Construct a feature vector containing day of the week, holiday, season, and weather conditions;

[0033] Input the feature vector into a trained time series prediction model to obtain time consumption estimates for each link of loading, transportation, and unloading;

[0034] Synthesize the time consumption of each link and combine the expected order time to calculate the expected arrival time.

[0035] As a further technical solution of the present application: the step S5 of analyzing historical KPI indicators and service evaluation data for the route includes:

[0036] Aggregating historical key performance indicators of the line, the indicators including one or more of on-time delivery rate, damage rate, customer complaint rate;

[0037] Mining historical text feedback of the line using sentiment analysis model to analyze negative topics in service evaluation;

[0038] Generating a comprehensive evaluation result containing service level score and text evaluation summary.

[0039] As a further technical solution of the application: the step S6 of generating a visual report includes the following steps:

[0040] Based on the line, warehouse, destination and time range, online analytical processing of shipping history data is performed to aggregate total tonnage, number of shipments and trend data;

[0041] Using a visualization library to automatically generate charts showing shipping trends and distribution statistics;

[0042] Based on the aggregated data, a text summary is generated using natural language generation technology.

[0043] As a further technical solution of the application: the step S7 of automatically detecting restriction policies and providing detour suggestions includes the following steps:

[0044] From a policy rule library updated regularly through a crawler or API interface, query restriction policies related to the origin and destination and transportation time;

[0045] If the predicted transportation time conflicts with the restriction policy, a path planning algorithm is called to recalculate the detour route and estimate the additional distance and time of the new route.

[0046] As a further technical solution of the application: the step S8 of simulating the prediction of each key time node in the transportation process includes the following steps:

[0047] Taking the predicted order placement time as the starting point, based on the time consumption estimation of each link, using discrete event simulation method to sequentially calculate the predicted time of order taking, dispatching completion, pickup completion, in-transit transportation, arrival at destination and unloading and signing;

[0048] Providing a confidence interval for each predicted time point;

[0049] Organizing the entire simulation timeline into a full-process prediction explanation in natural language form and outputting it to the user.

[0050] As a further technical solution of the application: all steps of the method are implemented through an AI dialogue agent integrated in an AI dialogue agent, which is based on a large language model to support natural language multi-round interaction with users.

[0051] In summary, compared with the prior art, the present invention has at least one of the following beneficial technical effects:

[0052] 1. This invention discloses a one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence. Through an AI-powered conversational logistics prediction system, users can conveniently obtain one-stop logistics information prediction services via natural language interaction. These services include intelligent warehouse recommendations for receiving and dispatching goods, route optimization suggestions, estimated loading / transportation / unloading times, intelligent calculation of freight and loading / unloading costs, reference for transportation unit prices, route service level assessment, regional climate information, and analysis of historical shipping data from the same region. The system's AI-driven conversational approach significantly lowers the barrier to entry, enabling relevant personnel to quickly grasp key information in the transportation process through natural dialogue, thereby facilitating advance decision-making and significantly improving logistics operational efficiency and user experience.

[0053] 2. This invention revolutionizes the traditional logistics information retrieval model by integrating natural language processing, multi-source data fusion, and intelligent prediction models. It liberates users from tedious, cross-system manual queries and information integration, allowing them to obtain one-stop, multi-dimensional decision support through a single dialogue interface. This greatly improves the efficiency and convenience of information acquisition, achieving a fundamental shift from "people searching for information" to "intelligent information delivery to people."

[0054] 3. This invention goes beyond simple information retrieval, achieving a leap from static retrieval to dynamic prediction. Its innovation lies in the comprehensive use of AI technologies such as time series prediction, neural network recommendation, and spatiotemporal simulation to accurately quantify and proactively simulate transportation costs, timeliness at each stage, key nodes, and even the impact of external policies. This generates deep insights that traditional methods cannot provide, enabling logistics decisions to shift from relying on experience to data-driven scientific prediction, significantly improving the accuracy and reliability of planning.

[0055] 4. Ultimately, this invention empowers enterprises to conduct forward-looking and precise logistics management. By providing high-precision cost forecasting, full-process visual simulation, and route service level assessment, it enables enterprises to mitigate risks in advance, optimize resource allocation, and formulate optimal transportation plans. This effectively reduces operating costs, improves transportation timeliness and customer satisfaction, and achieves significant technical effects in enhancing the resilience and competitiveness of the overall supply chain. Attached Figure Description

[0056] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0057] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0058] Example 1:

[0059] Reference Figure 1 The present invention discloses a one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence, comprising the following steps:

[0060] Step S1: Identify key information in the user input and automatically match it with the line information that meets the conditions that has been maintained in the system;

[0061] Step S2: Intelligently filter and recommend suitable shipping warehouses through dialogue interaction;

[0062] Step S3: Call the preset price library to automatically calculate the unit price of transportation, special fees and total freight cost, and provide feedback in natural language.

[0063] Step S4: Based on the matched route information, intelligently analyze historical loading, transportation and unloading time data to generate an overall transportation time estimate;

[0064] Step S5: Analyze the historical KPI indicators and service evaluation data of the route using AI;

[0065] Step S6: Automatically query and generate a visual report of historical cargo shipment statistics, such as total tonnage and number of shipments;

[0066] Step S7: Automatically detect traffic restriction policies in the delivery and receipt areas and provide detour suggestions;

[0067] Step S8: Use AI simulation to predict key time nodes in the transportation process, such as order acceptance, scheduling completion, pickup completion, in-transit transportation, arrival at the destination, and unloading and signing, and provide users with a full-process prediction explanation in a dialogue manner.

[0068] Step S1, identifying key information in user input, includes the following steps: using a natural language understanding model to identify user intent and determine that the user intent is logistics query and prediction; using a named entity recognition model to extract key entities from user input, including destination, shipping tonnage, time information, cargo type and service requirements; mapping and standardizing the key entities with the main database inside the system to generate structured query conditions.

[0069] Specifically, in step S1, the input is a query request provided by the user in natural language (e.g., "I want to ship 10 tons of general cargo from the Shanghai warehouse to Haidian District, Beijing. I will place the order tomorrow afternoon and request delivery before next Friday") or in the form of a structured form.

[0070] The technical process is as follows:

[0071] 1. Natural Language Understanding (NLU):

[0072] Intent Recognition: First, a classification model (such as a Transformer-based model) is used to determine the user's intent as "logistics query and prediction".

[0073] Entity Extraction: Utilizing sequence labeling models (such as BERT-CRF) to accurately extract key entities from user statements, including but not limited to:

[0074] Origin: such as "Shanghai warehouse" (needs to be further mapped to the internal warehouse code of the system).

[0075] Destination: such as "Haidian District, Beijing" (address standardization and geocoding are required to resolve to specific geographical coordinates or administrative division codes).

[0076] Goods information: such as "10 tons" (weight) or "general cargo" (goods type, mapped to the category code defined by the system).

[0077] Time information: such as "tomorrow afternoon" (parsed into a specific date and time stamp YYYY-MM-DD HH:MM:SS) or "before next Friday" (parsed into the latest expiration time stamp).

[0078] Service requirements: such as implicit "standard shipping" or explicit "express" or "requires unloading" (mapped to Boolean values ​​or service level codes).

[0079] 2. Data Normalization and Mapping: The extracted entity values ​​are mapped and standardized with the main database inside the system (such as warehouse database, address database, product database) to ensure the consistency of query conditions.

[0080] 3. Line matching:

[0081] The standardized query criteria (such as origin and destination, cargo weight, product category, and time constraints) are used as input and matched with the route knowledge base.

[0082] Each route in the knowledge base defines attributes such as reachable area, weight range, category restrictions, default delivery time, and base fare.

[0083] The matching algorithm can use rule-based filtering (e.g., the destination must be within the delivery range and the weight of the goods does not exceed the weight limit) combined with vector similarity retrieval (used to handle all routes that match "Haidian District" to "Beijing") to filter out all potential transportation routes that meet the conditions.

[0084] Output: Generates a list of one or more matching line IDs, and assigns an initial matching score to each line. This list will serve as the basis for all subsequent steps.

[0085] Step S2, which involves intelligently filtering and recommending suitable shipping warehouses, includes the following steps: Based on the user-specified origin information, select available warehouses from the warehouse master data; Calculate a recommendation score for each available warehouse using a neural network recommendation model, taking into account multiple dynamic and static factors such as real-time warehouse capacity, historical shipping timeliness, current transportation resource shortage, and transportation costs; Output one or more warehouses with the highest recommendation scores and their reasons for recommendation.

[0086] Specifically, in step S2, the input includes: the user-specified departure point (e.g., "Shanghai"), the route list generated in S1, and real-time operation data.

[0087] The technical process is as follows:

[0088] 1. Initial warehouse screening: Based on the user-specified origin (e.g., "Shanghai"), filter all available warehouses located in that region from the system's warehouse master data.

[0089] 2. Multi-factor collaborative recommendation:

[0090] Construct a warehouse recommendation model (such as a gradient boosting decision tree (GBDT) or a deep neural network (DNN) that comprehensively considers the following dynamic and static factors to calculate a recommendation score for each available warehouse:

[0091] Static factors: distance between the warehouse and the starting point of the route, warehouse hardware facilities (such as whether they support loading and unloading of this type of goods), and warehouse fixed capacity.

[0092] Dynamic factors: the warehouse's real-time capacity (whether it is overloaded), the historical average delivery time to the destination, the current tightness of transportation resources (such as the number of available vehicles), and the current transportation cost to the route.

[0093] Business rules: Whether it meets the customer's contractual agreement regarding warehouse conditions and whether it belongs to the preferred warehouse network.

[0094] 3. Interactive Clarification: If the shipping location information entered by the user is ambiguous (such as only saying "shipped from East China warehouse"), or if the system determines that multiple warehouses are applicable but each has its own advantages and disadvantages based on the real-time status, the AI ​​assistant will clarify through dialogue (such as: "We recommend Shanghai Puxi warehouse and Suzhou warehouse for you. Puxi warehouse is 2 hours faster but the cost is slightly higher. Which one would you prefer?").

[0095] Output: Recommend 1-3 optimal shipping warehouse IDs and the reasons for the recommendation (e.g., "We recommend using the Shanghai Puxi warehouse, which is expected to allow for loading half a day earlier").

[0096] Step S3, which calculates the transportation cost, includes the following steps: based on the matched route and cargo information, query the benchmark freight unit price in the preset price library; based on the service requirements, calculate one or more special fees through the rule engine, including express fees, unloading fees, road and bridge fees, or fuel surcharges; and calculate the total transportation cost according to the formula: total cost = (benchmark unit price * chargeable weight) + sum of all additional fees.

[0097] Specifically, in step S3, input: the determined route ID, the shipping warehouse ID, detailed product information (weight, volume, category), and service requirements (expedited processing, unloading, etc.).

[0098] The technical process is as follows:

[0099] Benchmark freight rate query: Based on the route ID and cargo weight, query the corresponding benchmark freight rate (RMB / ton or RMB / cubic meter) from the price knowledge base.

[0100] Special fee calculation:

[0101] Additional Fee Calculation: Various additional fees are calculated based on the rules engine. For example:

[0102] Expedited Fee: If "Expedited or not" is true, a fixed fee or percentage fee will be charged according to the expedited rules table.

[0103] Unloading fee: If "Unloading required" is true, then refer to the unloading fee rate table based on the type and tonnage of the goods.

[0104] Special category surcharges: such as dangerous goods and refrigerated goods, may incur additional rates.

[0105] Road and bridge tolls and fuel surcharges: calculated dynamically based on route distance and current fuel price policies.

[0106] Total price calculation: Use the formula: Total cost = (Base unit price * Billable weight) + Sum of all additional charges. The billable weight may be the greater of the actual weight and the volumetric weight.

[0107] Natural Language Generation (NLG): Transforms computational results into easily understandable text. For example: "The total estimated freight cost is ¥5,200. This includes: basic freight ¥4,500 (15 tons * ¥300 / ton), express service fee ¥500, and unloading fee ¥200."

[0108] Output: A structured list of cost details and a cost description in natural language.

[0109] Step S4, generating the overall transportation time estimate, includes the following steps: extracting historical time-delivery data for matching routes from the historical waybill database; constructing feature vectors that include weekdays, holidays, seasons, and weather conditions; inputting the feature vectors into a trained time-series prediction model to obtain time estimates for each stage of loading, transportation, and unloading; synthesizing the time for each stage and combining it with the expected order placement time to calculate the expected arrival time.

[0110] Specifically, in step S4, input: route ID, shipping warehouse ID, destination, order time, and product type.

[0111] The technical process is as follows:

[0112] 1. Data Acquisition: Extract detailed timeliness data of all completed orders for this route within a recent period (e.g., the past 180 days) from the historical waybill database, including: warehouse order acceptance time, dispatch completion time, vehicle arrival time, loading completion time, transit time, arrival time at destination, and unloading completion time.

[0113] 2. Feature Engineering: Construct features for prediction, such as: day of the week, whether it is a holiday, season, weather conditions (obtain historical weather by accessing third-party APIs), product type, and shipping warehouse.

[0114] 3. Time Series Prediction Model: This model uses machine learning regression models (such as LightGBM and XGBoost) or deep learning sequence models (such as LSTM and Transformer) as the training set, learning the complex relationship between the aforementioned features and the time consumption of each stage (loading time, transportation time, and unloading time). The model can output an estimated time range (e.g., transportation time: estimated 38-45 hours) or a probability distribution for each stage.

[0115] 4. Overall Time Efficiency Calculation: The estimated time for each step is added together, and the buffer time between steps is considered to obtain the total transportation time. This is then combined with the user's estimated order time to calculate the estimated arrival time.

[0116] Output: Time estimates for each stage (e.g., loading: 2h, transportation: 40h±2h, unloading: 3h) and total estimated arrival time (e.g., expected delivery before 4 pm on October 27, 2023).

[0117] Step S5 involves analyzing the historical KPI indicators and service evaluation data for the route, including: aggregating and calculating the historical key performance indicators of the route, which include one or more of the following: on-time delivery rate, cargo damage rate, and customer complaint rate; using a sentiment analysis model to mine historical text feedback of the route and analyze negative themes in service evaluations; and generating a comprehensive evaluation result that includes service level scores and text evaluation summaries.

[0118] Specifically, in step S5, input: Line ID.

[0119] The technical process is as follows:

[0120] 1. KPI Calculation: Multiple key performance indicators for this route are calculated by aggregating data from historical data:

[0121] Time-sensitive categories: average transit time, on-time delivery rate.

[0122] Quality-related: damage rate, discrepancy rate.

[0123] Service category: Customer complaint rate, positive review rate.

[0124] Cost category: Cost per ton-kilometer.

[0125] 2. Sentiment Analysis: Use sentiment analysis models (such as BERT-based text classifiers) to mine historical customer service work orders and customer feedback texts for this route, and analyze the themes of negative reviews (such as "poor driver service attitude" and "delayed delivery").

[0126] 3. Visualization and Scoring: Transform KPI data into intuitive dashboards or scorecards (e.g., Overall Service Score: 92 / 100). The system can automatically generate a summary evaluation (e.g., "This route has a historical on-time rate of 95%, but there have been 3 complaints about rough handling in the past month. Please be aware.").

[0127] Output: Service level rating of the line, key KPI data charts, and a text evaluation summary.

[0128] Step S6, generating a visualization report, includes the following steps: based on routes, warehouses, destinations, and time ranges, online analysis and processing of historical shipping data are performed to aggregate total shipping tonnage, number of shipments, and trend data; a visualization library is used to automatically generate charts displaying shipping trends and distribution statistics; and based on the aggregated data, a text summary is generated using natural language generation technology.

[0129] Specifically, in step S6, input: route ID, shipping warehouse ID, destination, and time range (which can be set to recent by default).

[0130] The technical process is as follows:

[0131] 1. Data Aggregation: Based on input conditions, execute OLAP (Online Analytical Processing) queries from the data warehouse and perform aggregation calculations.

[0132] Total tonnage shipped, total number of shipments, and average shipment weight.

[0133] Daily / weekly / monthly shipping trends.

[0134] Distribution of major shipping customers and distribution of major product types.

[0135] 2. Visualization Generation: Automatically generate charts using visualization libraries (such as ECharts, D3.js):

[0136] Trend chart: Shows how shipment volume changes over time.

[0137] Bar charts / pie charts: Show the percentage of customers or products.

[0138] Summary panel: Displays key statistics.

[0139] 3. Natural Language Summary: The NLG module generates a text summary based on the data, such as: "In the past three months, this route has shipped a total of 450 orders, totaling 12,000 tons of goods, with an average monthly growth rate of 5%."

[0140] Output: A visual analysis report containing charts and text summaries (can generate images or HTML snippets).

[0141] Step S7, which automatically detects traffic restriction policies and provides detour suggestions, includes the following steps: querying traffic restriction policies related to the place of receipt and delivery and transportation time from a policy rule library that is updated regularly through web crawling or API interface; if the predicted transportation time conflicts with the traffic restriction policy, calling the route planning algorithm to recalculate the detour route and estimating the additional distance and time of the new route.

[0142] Specifically, in step S7, input: place of origin, destination, and planned transportation time.

[0143] The technical process is as follows:

[0144] 1. Policy Knowledge Base: The system maintains a policy rule base, which is updated regularly from government and traffic management department websites through web crawlers or API interfaces to provide local traffic restriction and prohibition policies (e.g., out-of-town trucks are prohibited from entering the ring expressway in a certain city from 7 to 9 a.m. every day).

[0145] 2. Policy Matching and Inference: Based on the origin and destination of transportation and time, the rule engine is used to perform matching queries in the policy database.

[0146] 3. Spatiotemporal route planning: If the predicted transportation time conflicts with the traffic restriction policy, the system calls a route planning algorithm (such as Dijkstra's algorithm or A* algorithm) to recalculate a route that avoids the restricted area and time period, and estimates the distance and additional time of the new route.

[0147] Outputs include policy reminders (e.g., "Warning: Your truck is expected to enter Beijing's Fifth Ring Road at 8:00 AM on Wednesday, during which time trucks from other areas are prohibited from passing.") and detour suggestions (e.g., "It is recommended to detour via the West Sixth Ring Road, which is expected to increase the distance by 35 kilometers and the travel time by approximately 40 minutes.").

[0148] Step S8 involves simulating and predicting key time nodes during the transportation process, including the following steps: starting from the expected order placement time, based on the estimated time consumption of each stage, the discrete event simulation method is used to sequentially calculate the predicted time for each node: order acceptance, scheduling completion, pickup completion, in-transit transportation, arrival at the destination, and unloading and signing; a confidence interval is provided for each predicted time point; and the entire simulation timeline is organized into a full-process prediction description in natural language and output to the user.

[0149] Specifically, in step S8, the inputs are: the time estimate for each stage generated in S4, the path planning result in S7, and the order placement time.

[0150] The technical process is as follows:

[0151] 1. Process Simulation Engine: Constructs a logical model based on discrete event simulation to simulate the entire transportation process.

[0152] Events: Order acceptance, dispatch, pickup, departure, on-the-way location update, arrival, unloading, and receipt.

[0153] Status: The status and timestamp of each event.

[0154] 2. Node Time Prediction: Starting from the order time, add the scheduling time, pickup time, and in-transit transportation time predicted by S4 to calculate the specific time of each key node.

[0155] Estimated scheduling completion time = Order placement time + Average scheduling time

[0156] Estimated pickup completion time = Estimated dispatch completion time + Vehicle arrival time at warehouse + Average loading time

[0157] Estimated arrival time = Estimated pickup completion time + Transportation time (adjusted for road conditions and weather)

[0158] Estimated delivery time = Estimated arrival time + Average unloading time

[0159] 3. Uncertainty expression: For each forecast time point, a confidence interval is provided (e.g., "Estimated pickup completion time: October 26, 14:00 ± 1 hour").

[0160] 4. Natural Language Narration: Organize the entire simulated timeline into a coherent, story-like narrative, and present it to the user through a dialogue interface.

[0161] Output: A structured shipping timeline, including the predicted times for all key milestones, and a complete natural language flow description (e.g., "Your order is expected to be received at 16:00 on the 25th, picked up before 14:00 on the 26th, arrive at its destination at 9:00 AM on the 28th, and be signed for before 11:00 AM.").

[0162] All steps of the method are implemented by integrating an AI dialogue agent, which is built on a large language model to support multi-turn natural language interaction with the user. This invention receives variable information provided by the user in natural language or structured form through AI dialogue interaction, which may include: destination address, shipping tonnage, estimated order time, shipping warehouse information, required delivery time, product type, whether expedited processing is needed, and whether unloading is required.

[0163] This invention employs an intelligent dialogue system based on a large language model. The large language model understands the logistics needs input by the user, including but not limited to: destination address, shipping tonnage, estimated order time, whether expedited processing is required, and whether unloading is required, among other multimodal information.

[0164] The system achieves the following core functions based on artificial intelligence technology:

[0165] 1. By employing intent recognition and entity extraction technologies, it automatically parses key information from user input and intelligently matches the optimal route;

[0166] 2. Employing a neural network recommendation algorithm combined with real-time transportation capacity data, the system intelligently filters and recommends the optimal shipping warehouse;

[0167] 3. The system dynamically calculates the unit price of transportation, special fees, and total freight costs through its price calculation function;

[0168] 4. Use time-series predictive neural networks to accurately predict the timeliness of each stage (loading, transportation, unloading) based on historical transportation data;

[0169] 5. Utilize optimized statistics to evaluate line service levels in real time and generate KPI indicators and service quality scores;

[0170] 6. Intelligently generate historical shipment statistics reports and trend forecasts;

[0171] 7. Using a spatiotemporal prediction model, simulate the entire transportation process time nodes, including key links such as order acceptance, scheduling, pickup, in-transit transportation, and arrival and signing.

[0172] The implementation principle of this invention is as follows: This invention discloses a one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence, belonging to the field of intelligent logistics technology. This method receives logistics demands from users through AI dialogue interaction using natural language or structured input, and uses intent recognition and entity extraction technologies to analyze key information; then, it intelligently matches the optimal transportation route, recommends shipping warehouses, and dynamically calculates freight costs; it comprehensively uses time-series prediction models to analyze historical data to estimate the timeliness of each stage, evaluates the service level of the route, and generates a visualized statistical report; it also automatically detects traffic restriction policies and provides detour suggestions; finally, it uses AI simulation to predict key time nodes throughout the entire transportation process. This method effectively solves the problems of scattered logistics information, reliance on manual labor, and inaccurate predictions, achieving efficient and accurate one-stop logistics decision support, and significantly improving the efficiency and intelligence level of logistics operations.

[0173] The embodiments described herein are preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape, and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence, characterized in that, Includes the following steps: Step S1: Identify key information in the user input and automatically match it with the line information that meets the conditions that has been maintained in the system; Step S2: Intelligently filter and recommend suitable shipping warehouses through dialogue interaction; Step S3: Call the preset price library to automatically calculate the unit price of transportation, special fees and total freight cost, and provide feedback in natural language. Step S4: Based on the matched route information, intelligently analyze historical loading, transportation and unloading time data to generate an overall transportation time estimate; Step S5: Analyze the historical KPI indicators and service evaluation data of the route using AI; Step S6: Automatically query and generate a visual report of historical cargo shipment statistics, such as total tonnage and number of shipments; Step S7: Automatically detect traffic restriction policies in the delivery and receipt areas and provide detour suggestions; Step S8: Use AI simulation to predict key time nodes in the transportation process, such as order acceptance, scheduling completion, pickup completion, in-transit transportation, arrival at the destination, and unloading and signing, and provide users with a full-process prediction explanation in a dialogue manner.

2. The one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence according to claim 1, characterized in that, The step S1 of identifying key information in user input includes the following steps: The user intent is identified using a natural language understanding model, and the user intent is determined to be logistics inquiry and prediction. Key entities, including destination, shipping tonnage, time information, cargo type, and service requirements, are extracted from user input using a named entity recognition model. The key entities are mapped and standardized with the main database inside the system to generate structured query conditions.

3. The one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence according to claim 1 or 2, characterized in that, The intelligent screening and recommendation of suitable shipping warehouses in step S2 includes the following steps: Based on the user-specified origin information, select available warehouses from the warehouse master data; Based on a neural network recommendation model, a recommendation score is calculated for each available warehouse. The recommendation score comprehensively considers multiple dynamic and static factors, including real-time warehouse capacity, historical delivery timeliness, current transportation capacity resource shortage, and transportation costs. Output one or more warehouses with the highest recommendation scores and the reasons for the recommendation.

4. The one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence according to claim 1, characterized in that, The calculation of transportation costs in step S3 includes the following steps: Based on the matched route and product information, query the benchmark freight unit price in the preset price library; Based on service requirements, one or more special fees are calculated using a rules engine. These special fees include express fees, unloading fees, road and bridge tolls, or fuel surcharges. The total freight cost is calculated using the formula: Total Cost = (Base Unit Price * Chargeable Weight) + Sum of All Additional Fees.

5. The one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence according to claim 1, characterized in that, The step S4 of generating the overall transportation time estimate includes the following steps: Extract historical timeliness data for matching routes from the historical waybill database; Construct feature vectors that include weekdays, holidays, seasons, and weather conditions; The feature vector is input into the trained time series prediction model to obtain the time estimates for each stage of loading, transportation and unloading. The estimated arrival time is calculated by combining the time consumed in each step of the process with the estimated order placement time.

6. The one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence according to claim 1, characterized in that, The analysis of historical KPI indicators and service evaluation data for this line in step S5 includes: The historical key performance indicators of the aggregated computing line include one or more of the following: on-time delivery rate, damage rate, and customer complaint rate; We used sentiment analysis models to mine historical text feedback on the routes and analyzed negative themes in service evaluations. Generate a comprehensive evaluation result that includes service level ratings and a summary of written evaluations.

7. The one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence according to claim 1, characterized in that, The step S6 of generating the visualization report includes the following steps: Based on routes, warehouses, destinations, and time ranges, online analysis and processing of historical shipping data are used to aggregate total shipping tonnage, number of shipments, and trend data. Use a visualization library to automatically generate charts that display shipping trends and distribution statistics; Based on aggregated data, text summaries are generated using natural language generation techniques.

8. The one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence according to claim 1, characterized in that, The automatic detection of traffic restriction policies and provision of detour suggestions in step S7 includes the following steps: Query traffic restriction policies related to the place of delivery and receipt and the time of transportation from a policy rule library that is regularly updated through web crawling or API interface; If the predicted transportation time conflicts with the traffic restriction policy, the route planning algorithm is invoked to recalculate the detour route and estimate the additional distance and time of the new route.

9. The one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence according to claim 1, characterized in that, The simulation and prediction of key time points during the transportation process in step S8 includes the following steps: Starting from the expected order placement time, based on the time consumption estimates of each stage, the discrete event simulation method is used to sequentially calculate the predicted time of each node: order acceptance, scheduling completion, pickup completion, in-transit transportation, arrival at the destination, and unloading and signing. Provide a confidence interval for each predicted time point; The entire simulation timeline is organized into a full-process prediction description in natural language and output to the user.

10. The one-stop intelligent retrieval and prediction method for logistics information based on artificial intelligence according to claim 1, characterized in that, All steps of the method are implemented by integrating an AI dialogue agent, which is built on a large language model to support multi-turn natural language interaction with the user.

Images